• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

6970 3dmark11 score

UtopiA said:
Except the 4890 was meant to compete with the 275, not the 285.
5870 vs 480 also isn't a good comparison, either. If the 6970 is between 10% and 20% faster than the 480 (assume 15% average) then it's a lot closer to the 580 than the 5870 was to the 480.

I... don't think there's any valid comparison for single GPU's.
Not in recent years, anyway. ATI has never been that close to Nvidia.
Click to expand...

I sort of see your point, but I don't agree entirely. The 5870 was and is pretty close to the GTX480, especially now that it's gotten quite a boost in games where it used to lag behind( Metro2033 for instance). It can't be more than 5-10% behind on average now( at 19*12 at least). I might have been wrong about the 4890 and 285, but I seem to remember those being rather close(15%'ish?) in performance too?

Do note, I'm not actually considering price or which card it's targeted to compete with at this point, just straight out fastest GPU performance. And I do realize that's not really what AMD/ATI have been aiming for the last few years, but it would be kind of cool to see them finally rise to take the throne of even the single-GPU market with what appears to be a really ingenious design philosophy. Which is why I'd be somewhat disappointed if this really just turned out like the last round.
 
I find it interesting we have so many "leaks" and yet not a single PCB or die shot with 5 days to go

Makes me wonder what's actually being leaked
 
5*General said:
They released the 68xx cards with malfunctioning MLAA drivers, so I doubt it's that.
Click to expand...

Well, that's just it.

There aren't any drivers for the 6870 really. Just had a few extra lines tacked into the .inf. For the next cards they need to get it right.
 
5*General said:
They released the 68xx cards with malfunctioning MLAA drivers, so I doubt it's that.
Click to expand...

What malfunction did you experience? You don't have a 6800 series card in your signature. I got my two at release and didn't notice any on either of my machines.
 
Posted these thoughts in the other thread:

FWIW, if 1408 and 1536 SP's are true, I can see why 3dMark / synthetic scores don't show AMD much faster than the last gen:

1408 * 800 * 2 = 2.252 TFlops
1536 * 880 * 2 = 2.703 TFlops

Now compare that to Cypress:

1440 * 725 * 2 = 2.088 TFlops
1600 * 850 * 2 = 2.720 TFlops

In terms of pure math output from shaders, Cayman is actually more or less even with Cypress. As most of you guys know, 3dMark utilizes shaders more effectively than real world gaming. The original VLIW5 shaders of Cypress could be utilized pretty effectively by 3dmark. Some people were saying that the 3dMark scores were ugly for AMD - granted, driver optimizations are key in this, but that's the big thing - 3dMark wasn't representative of real world performance, because its easy to optimize against benchmarks *cough* Nvidia *cough*

The move from VLIW5 -> VLIW4 is supposedly because VLIW4 in real world situations gets used more efficiently, to where VLIW4 and VLIW5 were near equivalent while being 10% smaller savings for size - but benchmarks can use the 5th shader in synthetics, and factors it into scores

The improved tesselation / TMUs / ROPS no doubt give Cayman the score boost over Cypress, particularly in 3dMark11, but in shader intensive benchmarks, it might actually do only even with Cypress

If you look at the Chinese translations, they more or less say the same thing - it loses in 3dMark but in real world gaming, it more than beats the 570 and keeps close to the 580

Of course, that's why I tend to ignore reviews with synthetic benchmarks and keep my reading to real world scenarios

Just some thoughts
 
ZerazaX said:
Posted these thoughts in the other thread:

FWIW, if 1408 and 1536 SP's are true, I can see why 3dMark / synthetic scores don't show AMD much faster than the last gen:
Click to expand...

Where do you get 1408 and 1536 from? The number I've been hearing consistently out of the FUD is 1920 SP for the 6970 which would be a 50% boost in shader performance. It makes no sense for AMD to release a 6900 series with brand new shaders that barely outperforms the 6870/5870 when it's clearly the flagship series.
 
vjcsmoke said:
Where do you get 1408 and 1536 from? The number I've been hearing consistently out of the FUD is 1920 SP for the 6970 which would be a 50% boost in shader performance. It makes no sense for AMD to release a 6900 series with brand new shaders that barely outperforms the 6870/5870 when it's clearly the flagship series.
Click to expand...

A lot of the latest rumors are 1536.
 
Last edited:
vjcsmoke said:
Where do you get 1408 and 1536 from? The number I've been hearing consistently out of the FUD is 1920 SP for the 6970 which would be a 50% boost in shader performance. It makes no sense for AMD to release a 6900 series with brand new shaders that barely outperforms the 6870/5870 when it's clearly the flagship series.
Click to expand...

The most credible rumors are 1536

1536/4 = 384 > 1600/5 = 320, and for real world gaming, the 1536 performs much better
 
Silus said:
So you need a certain card to know that AMD released several hotfixes, ever since the HD 6800s release, to fix several MLAA issues (among other things of course) ?

http://www.hardwarecanucks.com/news/video/amd-releases-catalyst-hotfix-1010e-update/

5 hotfixes in total!
Click to expand...

5 hotfixes are a hell of a lot better then waiting 5 months while the WHQL certification takes for bloody ever on each version. its pretty much why there hasnt been an official driver yet. considering the 10.11 drivers are technically 10.10b drivers or what ever, it just shows you how long it actually takes. its the same way with nvidia. their beta drivers are actually the drivers waiting for WHQL certification.

its better to get everything ironed out right now with the 6800 series with the hotfixes so that everything can be combined into a single WHQL driver thats most likely going to be released on launch day with the 6900 series.


heflys20 said:
Wrong topic!
Click to expand...

lol there really isn't a topic here. we already know the 3dmark numbers are garbage.
 
ZerazaX said:
The most credible rumors are 1536

1536/4 = 384 > 1600/5 = 320, and for real world gaming, the 1536 performs much better
Click to expand...

What makes 1536 more credible though?

Okay then, please try to make sense of this leaked slide showing the 6990 with 3840 shaders.
http://www.hexus.net/content/item.php?item=27626

Simple math. 3840/2=1920
How is AMD going to make a dual gpu card with 3840 shaders if no 1920 shader part exists?

Here's another mathmatical anomaly. Most of the so called leaked 'shots' of the 6900 series gpu Cayman have 1600 unified shaders shown by GPU-Z. As we both know that's BS. The 5870 has 1600 shaders, and neither the 6950 nor 6970 have ever been rumored to be slated for 1600. Therefore the actual shader count of the 6970 is unconfirmed by any shred of real evidence.

The other thing that makes zero sense is market differentiation. If the 6970 turns out to only be as fast as a GTX 570, then what is the 6950 useful for? The 6870 has already usurped the 470's place as the best ~240 part. The 6950 would be stuck in limbo.

More math thoughts. If we posit:
6970 - 1920
6950 - 1536

We have a reduction of exactly 20% shaders. And a clear differentiation in products.

Now let's hypothesize:
6970 - 1536
6950 - 1408 (BTW I've never heard this interpolated 1408 number leaked by any source so it appears highly dubious)

Now the difference in shaders is only 10%. Why would ANYBODY pay significantly more over the 6950 for the 6970 for maybe 10% of performance difference? There's way too little product differentiation.

IMO if the 6970 can't compete with the 580 or at least beat the 480/570 by a significant margin, AMD is in BIG trouble. And I just don't think AMD is dumb enough to release a product like that.

GTX 570 performance is pretty much identical to GTX 480 performance. With a new product line why would AMD target to only MATCH GTX 480 performance? That's the part that makes no sense. AMD went for the jugular with the 6800 series and blew the 460 out of the water and even ate into the 470's market share forcing Nvidia to slash prices by $100. I see no reason for AMD to simply try to 'match' the 480's known performance with their 6900 series when they were so aggressive in trying to beat the pants off the 460 and 470.
 
Because all the recent leaks and GPU-Z fix have it at 1536

And as to why 3dMarks cores are garbage compared to real game benches:

130253tz0ffiyt7tfec8ee.jpg
 
Really? I've never seen a GPU-Z screenshot that shows 1536 cores. They all show 1600 cores which means GPU-Z doesn't know what the hell the video card's actual specs are yet.

OgAAAFU6OfBnkoC17Kb47Apa0VY2o3zyA7vyOpzF4VZq1Ai8xCCsZ4pcL2VsvnDcsytjU_VYcSx3ldvq1_rm7v4AatQAm1T1UHZ5P1_xy1ssLoksxjQE8dbT33mw.jpg


Care to check your eyes next time?
 
vjcsmoke said:
Really? I've never seen a GPU-Z screenshot that shows 1536 cores. They all show 1600 cores which means GPU-Z doesn't know what the hell the video card's actual specs are yet.

OgAAAFU6OfBnkoC17Kb47Apa0VY2o3zyA7vyOpzF4VZq1Ai8xCCsZ4pcL2VsvnDcsytjU_VYcSx3ldvq1_rm7v4AatQAm1T1UHZ5P1_xy1ssLoksxjQE8dbT33mw.jpg


Care to check your eyes next time?
Click to expand...

Wrong! its 1536, there are benchmarks already out too :). I to create a new post w/ actual results.
 
Anywho, same guy stated in his benches:

GTX 480 / HD6970

RE5
101 / 109

F1 2010
56 / 68

3DMark11X
X1650 / X1798

3DMark11P
P5017 / P4939

Heaven2.1
687 / 644
Click to expand...

Also, for reference, the Computemark score of 454 sounds bad compared to the 580

http://h-5.abload.de/img/6970computemarkxmc9.jpg

But look at what the 5870 scores:

http://www.geeks3d.com/20100606/gpu-computing-directcompute-computemark-2-1-gtx-480-vs-hd-5870/

238!

The 6970 scored 90.7% higher than the 5870 in the same compute bench!
 
ZerazaX said:
Anywho, same guy stated in his benches:



Also, for reference, the Computemark score of 454 sounds bad compared to the 580

http://h-5.abload.de/img/6970computemarkxmc9.jpg

But look at what the 5870 scores:

http://www.geeks3d.com/20100606/gpu-computing-directcompute-computemark-2-1-gtx-480-vs-hd-5870/

238!

The 6970 scored 90.7% higher than the 5870 in the same compute bench!
Click to expand...

Which begs the question - is Fermi really a Beast or is the 6970 just a clunker? Can't wait for some solid data instead of these 'mysterious leaks' from germany.
 
vjcsmoke said:
Which begs the question - is Fermi really a Beast or is the 6970 just a clunker? Can't wait for some solid data instead of these 'mysterious leaks' from germany.
Click to expand...

Fermi is a computational beast, but that doesn't translate to gaming. And the Fermi score is actually at a different resolution if u take note
 
Vantage High

GTX 480 = H13354

HD6970 = H14743

Seeing as how the 5870 received closer to X1000 more points over its life from driver optimizations, safe to say 6970 has a lot more under the hood than people think
 
Comments about EACH OTHER end now. If you want to talk about the cards, great, but if you can't stay civil we can arrange a week off for you to observe and learn.
 
You must log in or register to reply here.
Back
Top